Hi bmabam, the problem is that NURBS solids are designed to be made up of large smooth surface sheets, and not a zillion little facets that are arranged to simulate a smooth surface which is what you are ending up creating with this method.

Each surface in a NURBS model has some more overhead to it than each polygon in a polygon mesh model. That's because of various things like trim curves being tracked separately from the underlying surface, and each trim curve has both a UV curve to it as well as a 3D curve.

When used as intended for having broad smooth surfaces, that extra overhead is not significant and in fact the data size of a NURBS solid made up of proper smooth surfaces will usually be a lot less than a polygon mesh equivalent, with the NURBS solid also being a lot more accurate as well. For example a NURBS sphere made up of 1 sphere surface is a lot less in data size then a polygon mesh that is arranged to simulate a sphere.

But I can't recommend this kind of an approach where you're ending up with a single little tiny plane surface in the NURBS solid for every single polygon in the polygon mesh. That will only work very well if you have a pretty simple polygon mesh like less than 2000 faces, or if the polygons actually describe the model well like if you have all boxes or something that is actually supposed to be made up of triangles like maybe a gemstone.

What you probably have is something that is actually supposed to be a smooth surface, but instead you have 100,000 little facets that are arranged in such a way to try and emulate a smooth surface. That is not a good kind of thing to have as NURBS.

Usually for such things to get a good result you need to reverse engineer broader smooth surfaces from the polygon mesh data. There is some specialized software that can do stuff like that, check out this previous post for some links to some of those reverse engineering programs:
http://moi3d.com/forum/index.php?webtag=MOI&msg=2469.3

- Michael